SPN

<h1>Soft Proposal Networks<br>for Weakly Supervised Object Localization</h1>

[Project] [Paper] [Supp] [More Resources]

Torch Implementation

The torch branch contains:

• the official torch implementation of SPN.
• the PASCAL-VOC point-based localization demo.

Please follow the instruction below to install it and run the experiment demo.

Prerequisites

• Linux (tested on ubuntu 14.04LTS)
• NVIDIA GPU + CUDA CuDNN
• Torch7
• Jupyter Notebook

Getting started

You can setup everything via a single command wget -O - https://git.io/v5wTS | bash or do it manually in case something goes wrong:

1. install the dependencies (required by the demo code):

   • loadcaffe
   • torchnet: luarocks install torchnet
   • optnet: luarocks install optnet

2. clone the torch branch:

   # git version must be greater than 1.9.10
   git clone https://github.com/ZhouYanzhao/SPN.git -b torch --single-branch SPN.torch
   cd SPN.torch
   export DIR=$(pwd)
   

3. install SPN:

   cd $DIR/install
   # install the GPU implementation of SPN.
   bash install.sh
   

4. download the PASCAL-VOC2007 dataset:

   cd $DIR/demo/datasets
   # trainval
   wget http://host.robots.ox.ac.uk/pascal/VOC/voc2007/VOCtrainval_06-Nov-2007.tar
   tar xvf  VOCtrainval_06-Nov-2007.tar
   # test
   wget http://host.robots.ox.ac.uk/pascal/VOC/voc2007/VOCtest_06-Nov-2007.tar
   tar xvf VOCtest_06-Nov-2007.tar
   

5. download pre-trained VGGNet model for fine-tuning:

   cd $DIR/demo/models/convert
   wget http://www.robots.ox.ac.uk/~vgg/software/very_deep/caffe/VGG_ILSVRC_16_layers.caffemodel
   # convert caffemodel to t7
   th convertVGG.lua
   

6. run the demo experiment:

   cd $DIR/demo
   bash ./scripts/Train_PASCAL.sh
   

7. visualize locating samples via demo/notebooks/vis.ipynb

Troubleshooting

If you run into error: identifier "THCudaBlas_Sgemv" is undefined during installation, update Torch7 to the latest version via cd <TORCH_DIR> && bash ./update.sh

PyTorch Implementation

Check here.

Caffe Implementation

coming

Citation

If you use the code in your research, please cite:

@INPROCEEDINGS{Zhu2017SPN,
    author = {Zhu, Yi and Zhou, Yanzhao and Ye, Qixiang and Qiu, Qiang and Jiao, Jianbin},
    title = {Soft Proposal Networks for Weakly Supervised Object Localization},
    booktitle = {ICCV},
    year = {2017}
}